Build better products with our product team.
What problem does this solve?Today every developer can see, edit, and publish every asset in your Agent Studio org. That forces teams to cap developer access to a small trusted group, blocking expansion into sensitive domains like HR, finance, and procurement. RBAC gives you fine-grained control over who can view, edit, and run each asset, so you can safely scale.What's included:Folders: Public (org-wide), Personal (drafts), and admin-created custom folders (e.g., "HR Integrations"). Assets inherit folder permissions automatically. Four roles: Manager, Developer, Operator, Viewer, assignable per user per folder or asset. Dependency enforcement: Running an asset requires Use permission on its full chain, including every action and connector it calls. Boundary states: When blocked, the UI names exactly what's missing and surfaces a "Contact owner" prompt.How to get accessThis feature is in Controlled Availability for a curated set of design partners. Key limitations in this release: no group-based permissions (individual users only), flat folder structure only, no bulk admin actions, and MCP Servers are outside RBAC scope.Docs: linked here
Connect Any MCP Server to Your Moveworks AI Assistant!Release Type: Net New capabilitySummary: If a system has an MCP server, you will be able to connect it to your Moveworks AI Assistant -- no custom integration work of plugin development required). Point it at your endpoint, configure auth, and your assistant can start calling its tools. For many of the apps you're already using like Linear, Notion, Slack, we're shipping preconfigured servers so setup takes seconds and your employees will be able lookup data and take actions immediately. For example: "Create a Linear issue and link it to this Notion doc" "Find all unread Slack threads where I'm mentioned and summarize them" "Draft a Notion page from my meeting notes and assign follow-ups in Linear"What problem does this solve? Every new tool integration used to require a custom build. MCP eliminates that — if a system has an MCP server, your assistant can call its tools with just an endpoint and auth config.How does it work?Navigate to MCP Workspace in Agent Studio Select a preconfigured server or add a custom one via endpoint URL + auth Your assistant discovers available tools automatically Users can immediately invoke those tools in conversationKnown Limitations: TBDDocsite Link: TBDMedia: TBDAdditional Detail:Preconfigured servers on launch will be:• Asana• Atlassian• Linear• Intercom• Monday• Notion• Sentry• ZapierYou can connect any MCP server (including Snowflake, Slack and ServiceNow), not just those on the pre-configured server list. The "pre-configured" apps above just support Dynamic Client Registration (DCR) which automates the setup.
We are deprecating the legacy admin experience in favor of the new admin experience launched in November 2025. Users have been redirected to the new experience for the past months. We anticipate no disruption to users.
ObjectiveEnable end users to control their Moveworks experience by allowing them to configure notification preferences and default user settings, improving relevance, adoption, and satisfaction. Problem StatementCurrently, Moveworks delivers notifications and responses in a largely one-size-fits-all manner.This can lead to: Notification fatigue Irrelevant alerts Limited personalization for global users (language, timezone, location)
Similar to Prevent Duplicate Tickets by Disabling Double Click, but in agent studio. In a conversational process, there is the check box to have the user confirm before proceeding. This results in Yes and No buttons where the bot is asking the user to confirm to move forward. On clicking Yes, the associated action is run. This button does not disappear and can be clicked multiple times allowing the action to be run multiple times. In my case, this is resulting in many duplicate tickets to be created. Please make some action like editing away the button, or not running the action a second time once it has been executed the once.
As of now, the only way to deal with attachments via agent studio is with the link to a Moveworks site where the file is uploaded. Moveworks instead should be able to handle attachments added to the chat itself instead of making people leave the chatbot to go to a new url.
When using user slots, 99% of the time they work fine. On occasion though they seem to return a list of the single user instead of a single user object even when the slot is not set to be a list. This causes the plugin to fail when trying to get a user attribute like their email address as it expects me to first get the user in the list. My workaround for this is to do the following expression to turn it into a list and then take the first entry:requestorEmailAddress: EVAL(): args: normalized_req: FLATTEN(): - data.requestor expression: normalized_req[0].email_addrHowever having to do this in every plugin and every place where a user object is being leveraged is a bit much.
Updating our Reasoning Engine to GPT 5.4.Read more about the GPT-5.4 upgrade to the Moveworks Reasoning Engine here.
Data Visualization for a better user experience
Feature request : Teams Channel resolver like we have for Slack It will be a huge advantage for us as our majority of customer using the Microsoft Teams
Current experienceWhen a user requests a chat with a live agent through the AI conversational layer, Moveworks transfers the last ~20 chatbot messages to the chat-handling platform (Salesforce in this case) and asks the user to provide a chat description. However, the transferred conversation may be outdated, unrelated to the current issue, and often lengthy without any summarization. This can create confusion for advisors, who must spend time determining whether the chat history is relevant and if it aligns with the user's description.Proposed enhancementBefore transferring a chat, use AI to assess the recency and relevance of previous conversations.If a related conversation is detected, provide a concise AI-generated summary of the issue and key discussion points instead of the raw transcript. Ideally, if sufficient context is available, the user should not be required to enter a separate description. If the previous conversation is not relevant (e.g., a different topic or several months old), do not transfer it as context. If no relevant conversation history exists, prompt the user to provide description as the primary context for the advisor.BenefitsFaster advisor ramp-up and reduced handling time Elimination of misleading or outdated context Better alignment between transferred context and the employee's current issue Improved employee and advisor experience More intelligent and seamless AI-to-human handovers
In Agent Studio, the plugins have positive and negative prompts. Currently we have to add each one one at a time. Can we please add in support to add them in bulk? I have 40+ plugins and will be doing prompt adjustment every month or so, so adding them all 1 at a time is a long exercise. Pasting in CSV of many prompts into positive or negative and it being parse correctly would help a lot.
Hello team,We would like to request a new enhancement to include an additional custom field “Estimated Resolution Time” in the MW default HR case notifications when an HR case is created and in the last activity notifications. Kindly let us know if you could take up this enhancement and implement the requested changes. We would appreciate your confirmation on the feasibility and the next steps.
We'll upgrade the summarizer model of Enterprise Search App summaries to be cleaner, more accurate and faster. This update will improve the formatting and structure of certain summaries so they're easier to read. It will also help search be more grounded in reality such as understanding what today's date is. Finally, the upgrade will improve the average load time of the summary so users see the summary with less waiting.
Having a Continuous Integration (CI) and Continuous Delivery/Deployment (CD) through automated pipeline to update Configuration changes across different Moveworks environment. Without it, it creates a lot of effort to re-create configuration from test environment(s) to production environment.
Add a Language Match Rate (LMR) metric to the MLS evaluation pipeline to directly measure language mismatch between queries and summaries, enabling per-language diagnostics across all 19 MLS languages.
Enhance Knowledge Studio to support HR Cases as a knowledge source by enabling it to ingest, analyze, and leverage case data from the HRSD module. This capability would allow Knowledge Studio to identify common employee inquiries, extract resolutions and best practices from resolved HR cases, and use that information to improve the accuracy and relevance of generated responses.
Invoke Now Assist-based AI Agents & other 3P AI agents over A2A.
Recently ServiceNow Studio has integrated dragging tabs within its studio interface. I”m looking to do the same on Agent Studio when doing development on conversational process, compound action, http actions, plugin. And I need to preview multiple tabs simultaneously.
Already have an account? Login
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.